@songsid/agend 2.1.4-beta.48 → 2.1.4-beta.49

package/dist/daemon.js

@@ -361,6 +361,18 @@ export class PendingWorkTracker {
 const NORMAL_ENTER_SETTLE_MS = 500;
 /** Bottom-ready re-poll cadence for Enter-dropping TUIs once the pane is quiet. */
 const BOTTOM_READY_POLL_MS = 250;
+/**
+ * Startup failed because the CLI's backend is unreachable (see backend-outage.ts).
+ * The session-id is deliberately KEPT; the fleet schedules a delayed retry.
+ */
+export class BackendUnreachableStartupError extends Error {
+    backend;
+    constructor(backend) {
+        super(`CLI failed to start: the ${backend} backend is unreachable — session kept for a later retry`);
+        this.backend = backend;
+        this.name = "BackendUnreachableStartupError";
+    }
+}
 /** Bounded wait (under the pane lock) for the prompt to return before retrying a dropped Enter. */
 const STRANDED_RETRY_READY_WAIT_MS = 30_000;
 /** Max "stranded text → submit → wait for prompt → re-check" rounds per delivery; each may send one recovery Enter. */
@@ -737,6 +749,7 @@ export class Daemon extends EventEmitter {
     runtimeIdentity;
     spawnGate;
     stormWindow;
+    backendOutage;
     logger;
     tmuxSessionName;
     tmux = null;
@@ -785,6 +798,7 @@ export class Daemon extends EventEmitter {
     resolveSpawnSettled = null;
     spawnDepth = 0;
     skipResume = false;
+    startupAborted = false;
     backgroundSessionRecoveryAttempted = false;
     /** Whether the last spawn started a fresh session (not resumed). */
     isNewSession = false;
@@ -1005,7 +1019,9 @@ export class Daemon extends EventEmitter {
         this.errorRecoveryDeadlineAt = 0;
         this.activeErrorPatternKey = null;
     }
-    constructor(name, config, instanceDir, topicMode = false, backend, controlClient, rootLogger, runtimeIdentity, spawnGate, stormWindow) {
+    constructor(name, config, instanceDir, topicMode = false, backend, controlClient, rootLogger, runtimeIdentity, spawnGate, stormWindow, 
+    /** Fleet-level "is this CLI's backend down?" memory — see backend-outage.ts. */
+    backendOutage) {
         super();
         this.name = name;
         this.config = config;
@@ -1016,6 +1032,7 @@ export class Daemon extends EventEmitter {
         this.runtimeIdentity = runtimeIdentity;
         this.spawnGate = spawnGate;
         this.stormWindow = stormWindow;
+        this.backendOutage = backendOutage;
         if (!rootLogger)
             throw new Error("Daemon requires a shared root logger");
         this.runtimeIdentity ??= {
@@ -1862,7 +1879,9 @@ export class Daemon extends EventEmitter {
                             this.stormWindow?.markRecovered(this.name);
                     }
                     catch (err) {
-                        this.logger.error({ err }, `Failed to respawn ${cliLabel} window`);
+                        if (!this.handOffBackendUnreachableRespawn(err)) {
+                            this.logger.error({ err }, `Failed to respawn ${cliLabel} window`);
+                        }
                     }
                 }
                 catch (err) {
@@ -2546,15 +2565,20 @@ export class Daemon extends EventEmitter {
         }
         this.pauseWakeState = "waking";
         this.beginSpawn();
+        // A wake is a real resume launch: it gets at least the backend's resume
+        // budget (kiro: 60s — the conversation must come back from the backend
+        // before anything paints), never less than the caller's or the configured
+        // startup_timeout_ms.
+        const budgetMs = this.wakeBudgetMs(timeoutMs);
         const transition = this.autoPauseController.wakeOnDeliver(async () => {
             let timeout;
             try {
                 const ready = await Promise.race([
-                    this.trySpawn(true, timeoutMs),
-                    new Promise(resolve => { timeout = setTimeout(() => resolve(false), timeoutMs); }),
+                    this.trySpawn(true, budgetMs),
+                    new Promise(resolve => { timeout = setTimeout(() => resolve(false), budgetMs); }),
                 ]);
                 if (!ready)
-                    throw new Error(`Wake timed out before CLI became ready (${timeoutMs}ms)`);
+                    throw new Error(`Wake timed out before CLI became ready (${budgetMs}ms)`);
             }
             finally {
                 if (timeout)
@@ -4671,9 +4695,35 @@ export class Daemon extends EventEmitter {
                 throw new Error("No backend configured — cannot spawn CLI window");
             }
             const attemptedResume = !this.skipResume;
-            const alive = await this.trySpawn();
+            // A resume launch may get a longer budget than a fresh one (kiro: the
+            // conversation must come back from the backend before anything paints).
+            const resumeBudget = attemptedResume ? this.startupBudgetFor(true) : undefined;
+            let alive = await this.trySpawn(false, resumeBudget);
+            if (!alive && attemptedResume) {
+                // Resume failed. Before abandoning the session:
+                //  1. If the backend is unreachable (fleet-level memory, or this very pane
+                //     shows the outage text), clearing the session would trade a
+                //     transient outage for permanent conversation loss — and the fresh
+                //     start would hang on the same backend. Fail the startup and keep the
+                //     session; the fleet's delayed retry comes back once it is up.
+                //  2. Otherwise, for backends that ask for it, retry resume ONCE — the
+                //     first miss is usually slowness, not a broken session.
+                await this.noteStartupPaneForBackendOutage();
+                await this.failStartupIfBackendUnreachable();
+                if (this.backend.retriesResumeOnStartupFailure?.() === true) {
+                    this.logger.warn("Resume startup failed — retrying resume once before abandoning the session");
+                    await this.killProcessTree();
+                    await this.tmux.killWindow();
+                    alive = await this.trySpawn(false, resumeBudget);
+                    if (!alive) {
+                        await this.noteStartupPaneForBackendOutage();
+                        await this.failStartupIfBackendUnreachable();
+                    }
+                }
+            }
             if (!alive) {
-                // First attempt failed (stale --resume, crash, rate limit, etc.)
+                // Resume (or a fresh start) failed for a reason we do not recognise as an
+                // outage (stale --resume, crash, rate limit, etc.).
                 // Clean slate: clear session-id, skip resume, and retry once.
                 this.logger.warn("CLI startup failed — clearing session-id and retrying without resume");
                 const sidFile = join(this.instanceDir, "session-id");
@@ -4684,7 +4734,7 @@ export class Daemon extends EventEmitter {
                 this.skipResume = true;
                 await this.killProcessTree();
                 await this.tmux.killWindow();
-                const retryAlive = await this.trySpawn();
+                const retryAlive = await this.trySpawn(false, this.startupBudgetFor(false));
                 if (!retryAlive) {
                     await this.killProcessTree();
                     await this.tmux.killWindow();
@@ -4693,6 +4743,9 @@ export class Daemon extends EventEmitter {
             }
             else if (attemptedResume) {
                 resumedSuccessfully = true;
+                // A resume needs the backend: its success is positive proof the backend
+                // is reachable again (a fresh prompt is local and proves nothing).
+                this.backendOutage?.clear(this.backendKey());
             }
             this.lastSpawnAt = Date.now();
             this.skipResume = false; // CLI started successfully — reset for next spawn
@@ -4703,6 +4756,122 @@ export class Daemon extends EventEmitter {
         }
         return resumedSuccessfully;
     }
+    /**
+     * Startup budget for this launch: the backend's override (resume-aware), never
+     * below a user-configured startup_timeout_ms; undefined = trySpawn's default.
+     */
+    startupBudgetFor(resume) {
+        const override = this.backend?.getStartupBudgetMs?.({ resume });
+        if (override == null)
+            return undefined;
+        const configured = this.config.startup_timeout_ms;
+        return configured != null ? Math.max(configured, override) : override;
+    }
+    /** Effective wake budget = max(caller timeout, backend resume override, configured startup_timeout_ms). */
+    wakeBudgetMs(timeoutMs) {
+        return Math.max(timeoutMs, this.startupBudgetFor(!this.skipResume) ?? 0, this.config.startup_timeout_ms ?? 0);
+    }
+    /** The fleet-level backend key this instance runs on (matches the lifecycle's backendOf). */
+    backendKey() {
+        return this.runtimeIdentity?.backend ?? this.config.backend ?? this.backend?.binaryName ?? "unknown";
+    }
+    /**
+     * After a failed launch, read the pane for the backend's fleet-wide outage
+     * text (e.g. kiro's `dispatch failure (timeout) … kiro.dev`) and record it.
+     * The lifecycle only learns about outages from RUNNING instances; during a
+     * post-update herd nothing is running yet, so the startup path must report
+     * what it sees or the first instance in would clear its session for nothing.
+     */
+    async noteStartupPaneForBackendOutage() {
+        if (!this.backendOutage || !this.backend || !this.tmux)
+            return;
+        const outagePattern = this.backend.getErrorPatterns?.().find(p => p.fleetWide && p.type === "network");
+        if (!outagePattern)
+            return;
+        try {
+            const pane = await this.tmux.capturePane();
+            if (!outagePattern.pattern.test(pane))
+                return;
+            const detail = this.resolveErrorMessage(pane, outagePattern);
+            this.backendOutage.record(this.backendKey(), this.name, detail);
+            this.logger.warn({ detail }, "Startup pane shows the backend outage text");
+        }
+        catch { /* capture failed — nothing to learn */ }
+    }
+    /**
+     * Fail the startup WITHOUT touching the session while the backend is known to
+     * be down. The failed window is torn down first: the alternative — a live but
+     * never-ready CLI left behind while the daemon reports `crashed` — would keep
+     * the health monitor from ever seeing a dead pane and retrying.
+     */
+    async failStartupIfBackendUnreachable() {
+        if (!this.backendOutage?.isActive(this.backendKey()))
+            return;
+        this.logger.warn("Backend unreachable — keeping the session and failing startup for a delayed retry");
+        await this.killProcessTree();
+        await this.tmux.killWindow();
+        throw new BackendUnreachableStartupError(this.backendKey());
+    }
+    /**
+     * Crash-respawn ran into the backend outage: the ordinary path would leave
+     * `crashed` + a paused monitor + nothing scheduled (and count it as a crash).
+     * Hand the instance to the fleet instead — the lifecycle stops this daemon
+     * and the fleet's delayed startup retry re-creates it later with the session
+     * intact. Returns true when someone took the hand-off.
+     */
+    handOffBackendUnreachableRespawn(err) {
+        if (!(err instanceof BackendUnreachableStartupError))
+            return false;
+        if (this.listenerCount("startup_backend_unreachable") === 0)
+            return false;
+        this.logger.warn({ backend: err.backend }, "Respawn blocked by backend outage — handing off to the fleet's delayed startup retry");
+        this.healthCheckPaused = true;
+        this.emit("startup_backend_unreachable", { name: this.name, backend: err.backend });
+        return true;
+    }
+    /**
+     * Dispose a Daemon whose start() rejected. The lifecycle registers a daemon
+     * only after start() resolves, so nothing else owns this object — and by the
+     * time the spawn fails it has already bound its IPC server (channel.sock),
+     * written daemon.pid and possibly created a window. Left alone, each failed
+     * start leaks one live-but-unreachable server handle (the next start unlinks
+     * the socket path and binds a new one), multiplied by the automatic startup
+     * retries. Idempotent, and fast: no graceful CLI quit, the CLI never reached
+     * its prompt.
+     */
+    async abortStartup() {
+        if (this.startupAborted)
+            return;
+        this.startupAborted = true;
+        this.freezeRuntimeMonitors();
+        this.pendingIpcRequests.clear();
+        try {
+            await this.killProcessTree();
+        }
+        catch { /* nothing running */ }
+        if (this.tmux) {
+            const windowId = this.tmux.getWindowId();
+            try {
+                await this.tmux.killWindow();
+            }
+            catch { /* window may not exist */ }
+            if (windowId)
+                this.controlClient?.unregisterWindow(windowId);
+        }
+        try {
+            await this.ipcServer?.close();
+        }
+        catch (err) {
+            this.logger.debug({ err }, "IPC server close failed during startup abort");
+        }
+        this.ipcServer = null;
+        for (const file of ["daemon.pid", "window-id"]) {
+            try {
+                unlinkSync(join(this.instanceDir, file));
+            }
+            catch { /* absent */ }
+        }
+    }
     /** Kill the entire process tree of the current tmux pane (CLI + MCP server). */
     async killProcessTree(signal = "SIGTERM") {
         if (!this.tmux)